Per-metric multiset digest of the wide tiles, and of cell_metric

The wide tiles are a pivot of cell_metric; a pivot that dropped or duplicated a row is invisible in a row count. Both sides are reduced with hash_query()'s order-independent fingerprint over (cell_id, val) for cells that HAVE the metric, so the comparison cannot be fooled by row order or row-group layout.

Usage

app_cell_tile_digests(con, dir)

Arguments

con

a DBI connection to the release

dir

the app/cell directory

Value

a data frame metric_key, tiles, cell_metric, ok